1. Dust estimation via the triple window IR (8.7µm, 10.8µm, 12.0µm) jose.prieto@eumetsat.int

2.  Dust particle 10 µm  Earth globe 10 Mm   From micro to mega, twelve orders of magnitude difference in size  10 12 kg in the atmosphere (10 -7 of atmospheric mass) = fill all lorries!  Disputed human contribution to global cooling (S.K. Satheesh, 2006)  Trace for atmospheric circulation (polluting cars)  Life vector (Saharan protozoa and bacteria to the Caribbean) Can a satellite see dust particles ?

3. On IR imagery, dusty air appears cool in contrast to the hot daytime land surface. At night, the thermal difference between the background and the dust lessens. On VIS imagery over water, dust is easy to note. Over land, however, the dust plume and dry surfaces look similar Better dust detection in the infrared? Consecutive days in Fuerteventura, January 2010 Click one of the four fields, the one with best contrast between free-surfaces and dust areas

4. 2004-05-13 13:00 UTC, 0.8 µm Same date and time, 10.8 µm Dust on visible and infrared Desert scene, Southern Sudan Dust reflects back solar energy to space Dusty air rises and cools down

5. DUST RGB composite: the strength of infrared for dust detection IR RGB composite based on channels at 8.7, 10.8 and 12.0 µm Solar RGB composite based on channels at 1.6, 0.8 and 0.6 µm

6. World Atlas of Atmospheric Pollution. Editor: R. S. Sokhi Aerosol and health Impact on: agriculture (fertile fields), climate (radiative balance), aviation (ash in routes)

7. Jun2000-May2001 Average aerosol NASA Earth Observatory Aerosol is more than dust Dust Marine salt Smoke (industrial carbon, biomass burn) Ash Pollen Ice crystals ?

8.  Infrared dust properties  Where you learn how cool dust really is  A model of atmospheric dust  Where you learn to distinguish high thin from low fat  Validation via AERONET  Where you learn that models can help your eyes  Mixed scenes: cloud and dust  Where you learn that dust associates with water  Conclusions  Where you learn that there is more dust on books than books on dust Contents

9.  Dust storms occasionally reach up to 1km | 5 km | 10 km height, and are as thick as 100m | 1km | 5 km  Dust optical depth is around 0.1 | 0.5 | 1 (average on land) or 3 | 10 | 50 for storms, in the visible range. Optical thickness in the IR is about 40% of those values.  Dust absorbs and scatters infrared radiation in the Mie | Rayleigh | optical region  Aerosol density average in the atmosphere 10 -7 kg/m3 ( equivalent of depth 0.1 | 1 | 3 ) Dust characteristics

10.  Dust storms occasionally reach 5 km height, frequently thicker than 1km  Dust optical depth around 0.5 (average on land) or 3 for storms, in the visible range. Optical thickness in the IR is about 40% of those values.  Dust absorbs and scatters infrared radiation in the Mie region  Aerosol density average in the atmosphere 10 -7 kg/m3 ( equivalent of depth 0.1)  Dusty air ~ AOD=1 ~ 1 mg/m 3 ~ 1 g/m 2 Dust characteristics Σabs Σscat 0.55µm section

11. Dust enhancement in a single IR channel (?) 2004 May 13 th 13:00 Meteosat 10.8µm colour-enhanced (left) and gray-enhanced (below) (280-293 K) 8.7 µm 10.8 µm 12.0 µm

12. (-19K, 5K)(-19K, 12K) (-7K, 12K) Ch9 (upper left) and three independent differences 10.8- 12µm 10.8µm 8.7- 10.8µm

13. The 10.8µm-12µm difference (vertical) Dust Ch 10.8µm

14. 8.7 µm 10.8 µm 12.0 µm Comparison of water cloud and dust in the IR window Low cloud Dust storms

15. 12.0µm 10.8µm :::::::::::::::::::::::::::::::::::::::::::::::::::: Emissivity=0.25 Emissivity=0.15 :::::::::::::::::::::::::::::::::::::::::::::::::::: Emissivity=0.90Emissivity=0.75 :::::::::::::::::::::::::::::::::::::::::::::::::::: Absorption + scattering efficiencies Abs+ scatter Thin dust < 0.5 absorbs more 10.8µm 12.0µm goes forward Thick dust > 1.5 emits more 10.8µm dust particle Ground contribution Dust contribution

16.  10.8µm radiation is more absorbed and more backscattered by dust than 12.0µm  For dust or ash, arc is inverted due to the thinner contribution layer at 10.8µm  10.8µm channel shows higher BT than 12µm for thick dust due to higher emissivity MSG Natural (solar) RGB composite 4-July-2003 10:00 UTC “Reversed arc” for dust scenes: Ch9-Ch10 versus Ch10

17. Exercise: plot 9-10 versus 10 Why is the brightness temperature difference Ch9-Ch10 turning positive for very thick dust layers?  Dust shows a higher emissivity at Ch9 than at Ch10  Water vapour condenses on dust and favours Ch9 emissivity over Ch10 emissivity  Dust has a scattering component, higher at Ch9 than at Ch10  Dust Ch10 signal comes from a thicker (and therefore warmer) top layer T Ex F NEx T NEx Ex: explains NEx: does not explain

18.  Infrared dust properties  Where you learn how cool dust really is  A model of atmospheric dust  Where you learn to distinguish high thin from low fat  Validation via AERONET  Where you learn that models can help your eyes  Mixed scenes: cloud and dust  Where you learn that life is impossible without water  Conclusions  Where you learn that there is more dust on books than books on dust Contents

19.  Dust tends to higher levels far from the source, decreasing in particle size  Decrease in 12.0µm BT due to height and dust thickness (and size and...) Ground branches Dust branch Dust model Ground dust source Dust Colder ground  200 km 

20. Model assumptions (limitations)  ( 32x32 surroundings): min T10.8-T12.0 < -1.3K  Empirical AOT estimates for channel saturation: 0.14----1.3----3.5----4.8  Uniform dust type  Dust in the pixel at a single height  Size not dependent on height  Ground temperature affected by thick dust above  Good results in areas 200 km across  Four result categories:  Dust-free (or low-level only, or night-time, or dry ground)  Only dust traces  Dust  Mixed with cloud

21. Green-red dotted curve for (Tground=295, Tdust=270) Cyan curves for Tg=310, and two values of Td=270 and 290 Thick dust cloud at low level can be confused with a thin layer high above Reduction of the ground temperature by dust screening the sun (‘thermal deficit’) Use channel difference 8.7µm – 10.8µm (negative for thin, positive for thick) Graphical analysis Ch 9-Ch10 Channel10 BT (Tg=310, Td=290) (Tg=310, Td=270) (Tg=295, Td=270) 295 K 310 K 270 K 290 K

22. Ground temperature Dust column to the ground Dust top temperature Scattering or absorption efficiencies at 12.0 µm Scattering or absorption efficiencies at 10.8 µm Arc sensitivity to increasing... Graphical analysis Ch 9-Ch10 Channel10 BT (Tg=310, Td=290) (Tg=310, Td=270) (Tg=295, Td=270) The arc shape depends on temperatures (dust top, ground, dust vertical extension) and The arc shape depends on efficiencies (dust composition, size, shape) The dip in the curve depends on relative weights of efficiencies at 10.8 and 12.0 µm

23. AOT=0.3 0.6 1.0 1.3 2.0 Real (blue dots, right h.s.) compared with simulated (green-red dots left h.s. and lines) scatterograms based on Tg=318 Td=272 Σ11=0.6, 0.3 Σ12=0.2, 0.25 Dust column down to 50% of that temperature difference Smaller arcs, higher in the scatterogram, indicate less temperature contrast (Tg – Td) Td=272 Td=276 Tg=302 Tg=318 Dust (Td) and ground (Tg) temperatures estimates

24. IR model operation T 0 Initial T 0 Initial T d T d Min difference Max T 10 in neighbours Min T 10 in neighbours Ch 9-Ch10 BTD Ch10 BT a b If slope=b, refresh T 0 If slope=a, refresh T d

25. AreaMin9-10 > Thre(Time)NO_DUST AreaVar < ThreUnifUNIFORM PixelAnalysis InContextTH>thres3R9 < -1DUST NO_CONVERGENCEMIXED-CLOUD TH>thres2R9 < 2DUST D79 3GROUND DustDownDUST TH>thres1R9 < 3DUST D79 3GROUND TD > ColdThresDUST CIRRUS DUST TRACES Decision tree yes no 1.Subjective verification against masks, images and news media: Done 2.Verification from other sources (AERONET, LIDAR): In progress 3.Inter-comparison with other methods (Solar): Not started

26. threshold ch9-ch10 < -1.3K AOT =1.7, strong depth threshold ch9-ch10 < -1.3K AOT =2.8, too strong depth Due to location of minimum threshold NOT < -1.3K AOT not calculated Graphical validation

27. Optical thickness, retrieved from IR Dust RGB 2010-05-08 12UTC Icelandic ash Model results are quite different from the RGB visual impressions, and add information in RGB non-pink areas with small dust depth No direct comparison with mass loading without ash density estimate Applicable to volcano ash courtesy of F Prata

28. Ground versus dust skill 21Mar2010 12UTC Meteosat-9 The IR model separates the dust areas from the ground dry areas Dust Rock Sandy ground IR model does not usually pick on rock or sand areas

29. Model fails for atmospheric inversions  Occasionally, during night, thermal inversions duct dust at high speed  Due to the thickness, no negative 10.8µm – 12 µm difference appears above the dust  However, negative differences appear over clear ground

30. Red= [-2K (10.8-12.0)µm...4K]=NoRed Green= [-15K (8.7-10.8)µm..0K]=NoGreen Magenta areas are typically dusty: neither necessary nor sufficient condition Inside magenta areas, darker (less green) pixels show a smaller difference c7-c9 which means higher AOD The threshold in the red component (-2K) is exceeded in large areas of the dust storms. Blue component is most of the time saturated over desert areas during day. Dust RGB NoBlue= [261K (10.8µm)...289K]=Blue 2010-03-21 12UTC, Saharian region

31. 2004-05-13 13:00 UTC, 10.8 µm 9 1 1 2 2 3 3 4 4 1: Thick high cloud 2: Broken low cloud 3: Ground, drier air towards 4 4: Dust cloud The cloud-to-dust spiral in the differences diagram

32. RGB compared with the IR-MODEL BT7-BT9 (pinkness) Reduced ground temperature under the thicker layer of dust AOD IR-MODEL discriminates significantly several (>4) levels of AOD RGB discriminates <4 levels Pinkness is not a direct measurement of AOD for high AOD

33. C9 strenght C7-C9, pinkness Tg - Td Channel versus model parameter: correlations AOD Correlations are stronger for AOD > 2 Ch9 strength (compared with the other channels) is not a good indication of AOD Ch7 – Ch9 is a better indication of AOD, still poor The thermal deficit retrieved by the model is reverse-correlated to AOD

34.  Infrared dust properties  Where you learn how cool dust really is  A model of atmospheric dust  Where you learn to distinguish high thin from low fat  Validation via AERONET  Where you learn that models can help your eyes  Mixed scenes: cloud and dust  Where you learn that life is impossible without water  Conclusions  Where you learn that there is more dust on books than books on dust Contents

35. 12.02E 13.22N, model on image at 12UTC: theta=0.6 31C-39C size=29 5.52E 22.77N, model on image: theta=0.16 40C-47C size=31 2.66E 13.53 N, model on image: theta=0.8 33C-42C size=14 -5.94E 13.28N, model on image: theta=1.9, 31C-42C

36. Validation based on ground measurements (AOD units) AEROMETIR-MODEL 0.90.631-39 C 29 µm 0.350.240-47 C31 µm 2.11.9 31-42 C  1.60.833-42 C14 µm  0.4NO DUST (too uniform) 0.1NO DUST 1.72.630-38 C 0.03NO DUST IR-MODEL is too sensitive to temperature at the arc minimum Ch7-Ch9 = PINK Ch9-Ch10 AOD

37. SAMPLE VALIDATION based on AERONET ground measurements  Good agreement (+/- 30%) over desert grounds  Over the ocean or islands, lack of model sensitivity due to insufficient temperature contrast, dust thinness or uniform background for neighbour calculation  Better match for coarse than for fine aerosol  No sample validation done so far for dust temperatures (heights), using ground temperature. This is essential for evaluation of the thermal deficit

39. Other validation source: Nowcasting SAF dust flag  For the ocean, day time: R1.6/R0.6 high, T12.0-T10.8 high, SD(T10.8-T3.9) smooth  For the ocean, night time: same IR, T8.7-T10.8 high  For continental surfaces, day time: not cold T10.8, smooth T10.8, filters for cloud Nowcasting SAF dust flag and Dust RGB 21-Mar-2010 12 UTC

40.  Infrared dust properties  Where you learn how cool dust really is  A model of atmospheric dust  Where you learn to distinguish high thin from low fat  Validation via AERONET  Where you learn that models can help your eyes  Mixed scenes: cloud and dust  Where you learn that dust tends to soak  Conclusions  Where you learn that there is more dust on books than books on dust Contents

41. Low level dust forming a dust wall in Niamey (courtesy of E. Kploguede)

42. 2008-03-23 11:30 UTC Meteosat Ch9 Ice cloud Dust over ground Dust-cloud interaction Dust over sea Land with different emissivities Dust over sea Dust over ground Ice cloud Land with different emissivities What is the ice temperature at the cloud boundaries? 265 K 275 K 285 K

43. Real (left h.s.) compared with simulated (right h.s.) scatterograms based on Tg=308 Td=266 Σ 8.7 =.35,.2 Σ 11 =.6,.3 Σ 12 =.2,.25 and ground emissivity 85% at 8.7µm Marks at optical-thickness third-units from the right ends 9-10 7-9 Less emissive branch More emissive ground at 8.7µm Value added by the channel 8.7µm 7-9 9-10

44. Dust-cloud interaction Cloud-dust index: 2*ch9 – ch7 – ch10

45.  Infrared dust properties  Where you learn how cool dust really is  A model of atmospheric dust  Where you learn to distinguish high thin from low fat  Validation via AERONET  Where you learn that models can help your eyes  Mixed scenes: cloud and dust  Where you learn that life is impossible without water  Conclusions  Where you learn that there is more dust on books than books on dust Contents

46. A model based on three infrared window channels provides a set of parameters for dust storm severity Tdust, Tground and Depth values are essentially derived from 10.8µm and 12µm Channel at 8.7µm provides refinement at the dust end of the curves. Not at the ground branch, due to uncertain ground emissivity The model validation against AERONET is satisfactory, but other validation tools (NWCSAF, LIDAR) are needed Conclusions

47. A pattern for surface cooling by dust and particle size profiles will improve the simulation of the observed radiances Particle size affects channel emissivity in a way to be learnt, usable to reduce the gap between expected and real radiances (residuals) Looking into the BT’s for dust mixed with water or ice will clarify the role of aerosols in cooling the atmosphere and inhibiting rain (or hurricanes!). Coupling IR technique with existing methods for solar channels will allow the simultaneous retrieval of surface albedo and aerosol optical depth A calibration against the solar technique will provide skill for the IR estimate, even during the night Outlook

48. List of referred events: 2004-05-13 12:00, Sudan and Saudi Arabia 2008-02-02 06:00, Saudi Arabia 2008-03-23 12:00, Libya 2009-03-28 18:00, Argentina THANKS FOR YOUR ATTENTION ! Fish Cross-over

49. Can you not think of a question? Just choose one from the following: 1. Why do you see “pink” areas in southern Africa frequently? Is there a diurnal temperature cycle? 2. What can we do in case of thermal inversions? Do channel diagrams help identify those situations? 3. How can we produce the scatterograms by ourselves?